Introduction

In the United States, the annual incidence of hip fractures is approximately 296,000 [9]. Approximately 50% of these fractures are extracapsular and are referred to as intertrochanteric or pertrochanteric fractures. Historically, a sliding hip screw (SHS) has been the preferred implant to stabilize these fractures [5, 13, 18, 21, 24, 25, 35]. However, intramedullary hip screw devices have gained popularity for stabilizing this fracture type [2, 4, 7, 8, 10, 12, 16, 17, 29, 31, 32].

Intramedullary implants have mechanical and theoretical clinical advantages in comparison to a SHS [24]. Owing to its more medial placement than a SHS side plate, the intramedullary nail (IMN) is closer to the mechanical axis of the lower extremity which decreases the bending moments on the implant. Potentially, the IMN can be inserted percutaneously with lower blood loss and less periosteal disruption than a SHS. However, Parker and Handoll evaluated studies comparing intramedullary hip screws and SHS; they concluded that given the increased risk of operative and later fracture of the femur and increased reoperation rate associated with IM hip screws, SHS are superior for treating intertrochanteric fractures [33].

Many of the published studies comparing SHS and IMN for stabilization of intertrochanteric fractures have involved a small number of patients and looked at a limited number of outcomes such as mortality and need for revision surgery [2, 4, 7, 8, 10, 12, 16, 17, 29, 31, 32]. Our study, using a large cohort of Medicare beneficiaries, was performed to determine whether patients who sustain an intertrochanteric fracture have better outcomes when stabilized with a SHS or an IMN. We compared 1-year revision surgery rates, mortality, length of hospital stay, number of days using rehabilitation services, and costs associated with both implants.

Materials and Methods

We used a 20% sample of Medicare claims (Part A and Part B) submitted by hospitals and physicians between 1999 and 2001 to identify a cohort of patients who had sustained intertrochanteric femur fractures. This 20% random sample was provided by the Center for Medicare and Medicaid Services (CMS) and was used in a previous study [23]. We identified 43,659 intertrochanteric femur fractures. The International Classification of Diseases, 9th Revision (ICD-9) code 820.2 listed as the primary diagnosis and the Current Procedural Terminology (CPT) codes 27244 and 27245 were used to identify two fracture groups defined by the implant chosen for fracture stabilization, an IMN or a SHS. Patients were excluded if they were an HMO participant, sustained a fracture secondary to cancer or major trauma (based on diagnosis codes found on the MEDPAR and relevant Part B claims at the time of index hospitalization), or had sustained a hip fracture during the previous calendar year. If a patient had more than one hip fracture during the study period, the first fracture was used in the analysis. Patients with conflicting ICD-9 diagnoses not specific for the intertrochanteric region (eg, subtrochanteric, femoral neck fracture) also were excluded. An SHS was used to stabilize 40,828 (94%) fractures and an IMN was used in 2831 (6%) fractures.

Demographic characteristics that were recorded included patient age (divided into five categories: 65–69, 70–74, 75–79, 80–84, 85 years or older), gender, race (black, not black), Medicaid status at the time of fracture (yes/no), disability as reason for Medicare status (yes/no), and hospital location (characterized in three groups based on rural-urban commuting codes [19]: urban, large town/suburban, small town/isolated). Hospital hip fracture volume was stratified on three levels that approximated terciles: low (less than 250 cases per year), middle (250–630 cases per year), or high (more than 630 cases per year) to adjust for volume-outcome effects. The demographics of the two implant groups were similar (Table 1). Patients in both cohorts were predominantly white, female, and older than 80 years. Chronic illness burdens before the hip fractures were similar for the two groups as per comparison of the number of comorbidities.

Table 1 General characteristics of the cohort
Full size table

We controlled for medical comorbidities using the Iezzoni comorbidity index, an accepted approach to comorbidity adjustment with administrative databases [20], using inpatient claims at the time of the index hospital admission and inpatient and outpatient claims during the 6 months before the index hospitalization. Diagnoses derived from Part B (physician services) claims were restricted to those defined as evaluation and management or procedures. Comorbidities were summed into the following categories: 0, 1, 2, or 3+ for analysis.

We used the denominator file to determine mortality at 30 days and 1 year after hip fracture. MEDPAR records were used to determine the length of hospital stay for the initial hospitalization and the 6-month period starting with the index admission. CPT codes (27090, 27236, 27130, 20680, 27132) were used to identify patients who required revision hip surgery secondary to a complication with fracture fixation within 1 year of the index procedure.

MEDPAR skilled nursing facility claims, Part B, and Home Health files were used to determine the total number of days using rehabilitation services for the first 6 months after hospital discharge. These included days spent in a skilled nursing facility, days with claims for home health or physical therapy, and comprehensive outpatient rehabilitation facility visits. Days in which patients had more than one type of therapy claim were not double-counted.

Total expenditures were calculated using standardized costs to allow comparisons free of payment differences resulting from region or teaching status of the hospitals; these costs included diagnosis-related group (DRG) weights for acute hospital admissions and total relative value units (RVU) for all Part B (physician supplier) bills. We evaluated the period from index admission to 1 year after injury (including the cost of the index hospitalization and procedure). Values were expressed in US dollars for 2000.

We used summary statistics (means, proportions) to describe the demographic profiles for the patients stabilized using either an IMN or SHS defined by ICD-9 and CPT codes. Logistic regression was used to predict the odds of binary outcomes (eg, death, revision surgery) and analysis of variance for continuous outcomes (costs, length of hospital stay, use of rehabilitation services). All models were adjusted (adjusted odds ration or AOR) for patient age, gender, race, RUCA categories, hospital fracture volume, Medicaid status, disability status, and number of comorbidities.

Results

Similar 30-day and 1-year mortality rates were found between patients treated with both implants (Table 2). The 30-day mortality rates for the IMN and SHS groups were 14.2% and 15.8%, respectively (AOR, 0.99; 95% confidence interval [CI], 0.89–1.11). One-year mortality rates for the IMN and SHS groups were 30.7% and 32.5%, respectively (AOR, 1.00; 95% CI, 0.90–1.07). The 1-year revision surgery rate was higher (AOR, 1.35; 95% CI, 1.16–1.57, p = .0001) in the IMN group compared with the SHS group (7.2% and 5.5%, respectively) (Table 2).

Table 2 Primary and secondary outcome measures by treatment
Full size table

The mean number of inpatient days for the index hospitalization was higher (0.17; 95% CI, .012–0.32; p = .035) for the IMN group than the SHS group (6.5 and 6.3 days, respectively) (Table 2). The mean numbers of days spent as an inpatient during the first 6 months after injury for the IMN and SHS groups were 9.6 days and 9.3 days, respectively, which was not statistically significant (0.17; 95% CI, −0.16–0.49).

The total mean numbers of days using rehabilitation services for the IMN and SHS groups were 10.2 and 9.5 days, producing a significant adjusted increase in the IMN group of 0.56 days (95% CI, 0.07–0.93, p = 0.015) (Table 2). Total costs (all physician and hospital costs) during the first year after injury were higher ($947; 95% CI, $581–$1313; p < .001) for patients in the IMN group ($16,854) than the SHS group ($15,710).

Discussion

This study was performed using a large cohort of Medicare beneficiaries to determine whether patients sustaining an intertrochanteric fracture have better outcomes when stabilized with a SHS or an IMN. We compared 1-year revision surgery rates, mortality, length of hospital stay, number of days using rehabilitation services, and costs associated with both implants.

The study’s limitations include the fact this is a retrospective database study with all the problems inherent with this methodology. Although the patient groups appeared similar, patients were not randomly assigned to one of the implant groups; thus, unmeasured confounders may exist that were not adjusted for in this analysis and could have biased the results. However, using a national database of Medicare claims should provide a true indication of the actual care that is being received across the breadth of institutions in the United States, whereas trials often are performed in select institutions using restrictive entrance criteria. Similar to most database projects, users cannot independently verify the accuracy of the data. However, Baron et al. performed internal validation of Medicare data for patients sustaining a hip fracture or undergoing prostatectomy, comparing hospital and physician claims [6]; they found excellent agreement between the two claim sources with percentage of agreement generally between 89% to 99%.

Furthermore, the Medicare database does not include detailed clinical information such as medications, severity of the associated comorbidities, lifestyle factors, body composition of the patient, or radiographic information such as fracture comminution or displacement. Two studies had high failure of fixation for reverse obliquity intertrochanteric fractures managed with a SHS and recommended use of fixed angled devices or IMN for this fracture configuration [15, 22]. Because the incidence of reverse obliquity fractures is not known for our cohort, the higher reoperation rate in the IMN group could have been influenced by inclusion of patients with greater fracture instability.

The study period was from 1999 to 2001. This period was relatively early during acceptance of the IMN for stabilization of intertrochanteric fractures as evidenced by the fact that 94% of our cohort was treated using a SHS. Although the demographic characteristics of the two groups were not significantly different, there may have been a learning curve for use of IMN for intertrochanteric fracture stabilization. This learning curve may have influenced the higher complication rate associated with use of the IMN. The results may have been different using a more recent cohort as clinicians became more familiar with use of IM implants for stabilization of proximal femur fractures.

Therefore, we do not have all information necessary to fully evaluate the appropriateness of specific interventions or to control for all relevant patient and surgeon factors that may have affected the complication rates after hip fractures. In addition, billing codes may not always completely reflect the details of the clinical care. There is no way to determine whether the implants used were short or long IMN; the complication rates between these two types of IMN might be different. Furthermore, the problem of laterality (ie, the fact that ICD-9-CM codes do not reliably distinguish between left and right sides) prevented us from knowing with certainty whether adverse events after hip fracture treatment were related to the extremity treated surgically at the index stay. However, this limitation would apply to patients treated with either implant.

Trochanteric antegrade IM nailing of intertrochanteric fractures using a large screw placed up the neck to proximally interlock a short IMN gained popularity in the 1980s and 1990s. Early reports suggested some advantages to this fixation technique, including a minimally invasive surgical technique, shortened operating times, lower blood loss, improved biomechanics, greater fracture stability, earlier mobilization, and shorter lengths of stay [7, 8, 11, 16, 26]. However, authors soon reported numerous technical complications, including fracture of the femur below the nail and need for revision surgery [7, 8, 11, 26]. The initial nails, made by several manufacturers, were redesigned with a smaller nail and locking bolt diameter and lower angle proximal bend.

Despite the initial problems associated with initial trochanteric nails for treatment of intertrochanteric fractures, there has been increased use of these implants. Using American Board of Orthopaedic Surgery (ABOS) Part II data, Anglen and Weinstein reported that use of IM fixation increased from 3% in 1999 to 67% of cases in 2006 [3]. Numerous studies have been published comparing sliding compression hip screw and side plate with IM fixation [2, 12, 16, 17, 29, 31, 32]. However, results have been contradictory in terms of outcomes [2, 12, 16, 17, 29, 31, 32] with the only consistent differences reported between the two fixation techniques being increased complications (particularly intraoperative and postoperative fractures) and a higher reoperation rate with IMN [4, 10, 16, 29].

After adjusting for relevant covariates, we found considerably worse results for the IMN procedure based on the rates of revision surgery, length of stay during the index hospitalization, number of days using rehabilitation services, and total costs accumulated during the year after fracture inclusive of the index hospitalization. We found no differences in the adjusted 30-day or 1-year mortality rate or number of days spent hospitalized during the first 6 months after fracture.

The risk for revision surgery during the first postoperative year was 35% greater for the IMN group compared with the SHS group (7.2% and 5.5%, respectively). This result is consistent with published revision surgery rates for intertrochanteric femur fractures managed with IMN and SHS. In a prospective, randomized trial of 400 intertrochanteric fractures randomized to either a Gamma nail (Stryker Medical, Mahwah, NJ) or SHS, Adams et al. reported 1-year revision surgery rates of 6% for the IMN and 4% for the SHS [1]. In a retrospective review of 921 pertrochanteric fractures treated in Oslo, Norway, Osnes et al. reported revision surgery rates of 11.7% for fractures stabilized using a Gamma nail and 8.1% for those treated with a SHS [30]. Parker and Handoll, comparing use of a cephalomedullary nail with a SHS for stabilization of extracapsular hip fractures, reported the short IMN was associated with a significantly greater revision surgery rate than a SHS (relative risk, 1.56; 95% CI, 1.12–2.18) [33].

The crude and adjusted 30-day and 1-year mortality rates for the two groups were not significantly different and the values reported in our analysis are consistent with published rates [14, 16, 27, 28, 33, 34]. In a prospective, randomized series of 100 intertrochanteric fractures stabilized with either a cephalomedullary nail or SHS, Hardy et al. reported a 1-year mortality rate of 30% using either implant [16]. Parker and Handoll, comparing use of a cephalomedullary nail with a SHS for stabilization of extracapsular hip fractures, reported no difference in mortality with use of either implant [33].

The difference in adjusted length of stay between the IMN and SHS groups for the index hospitalization and during the first 6 months after injury was 0.17 days. This difference was statistically significant only for the index hospitalization as a result of the greater variability in days of hospitalization during the 6-month period, but was not of much clinical importance. These results are interesting considering the higher revision surgery rate for fractures stabilized using an IMN; because of this higher revision rate, one would expect a greater number of hospital days for patients treated with the IMN. The IMN group had an additional 0.5 day using rehabilitation services during the first 6 months after injury. This difference represents an increase of only approximately 5%. As a result of the subjective nature of treatment plans, this small difference could easily be caused by differences unrelated to the type of surgical treatment used.

Total standard costs were higher for the IMN group in unadjusted and adjusted analyses. Patients managed with IMN had $332 and $617 higher adjusted costs for hospital reimbursements and Part B claims during the first year, respectively. Overall adjusted cost estimates, combining the two (DRGs + RVUs) were $947 higher for the IMN group than for the SHS group. This represents an increase of approximately 6%. The higher costs for the IMN group can be explained partly by the higher rate of revision surgery, longer length of stay, and higher physician RVUs associated with using an IMN compared with a SHS (20.31 and 15.94 RVUs, where 1 RVU = $36.61). The average increased spending of $950 per patient during the first year after fracture becomes financially important in light of the the current 296,000 annual hospital admissions for patients with hip fractures and the anticipated increase in the number of future hip fractures [9].

Our study confirms the findings of others regarding a higher revision surgery rate for intertrochanteric femur fractures stabilized with an IMN compared with a SHS. In light of the higher revision surgery rate and total expenditures for the IMN group, this analysis does not support routine use of an IMN for management of all intertrochanteric femur fractures.